ameliorating brain image segmentation using fuzzy ... · ameliorating brain image segmentation...

International Journal of Science and Research (IJSR) ISSN (Online): 2319-7064

Index Copernicus Value (2013): 6.14 | Impact Factor (2013): 4.438

Volume 4 Issue 5, May 2015

www.ijsr.net Licensed Under Creative Commons Attribution CC BY

Ameliorating Brain Image Segmentation Using

Fuzzy Clustering Techniques

Sana Tak1, Toran Verma

2

1M.Tech Scholar, Computer Science & Engineering Department,

Rungta College of Engineering and Technology, Kohka Kurud Road, Bhilai, Chhattisgarh, India

2M.Tech Co-ordinator, Computer Science & Engineering Department,

Rungta College of Engineering and Technology, Kohka Kurud Road, Bhilai, Chhattisgarh, India

Abstract: Image segmentation is usually outlined as a partition of pixels or image blocks into undiversified teams. These teams area

unit characterized by a prototypic vector in feature area, e.g., the area of Gabor filter responses, by prototypic histograms of options or

by combine wise dissimilarities between image blocks. For all three information formats price functions are planned to live distortion

and, thereby, to cipher the standard of a partition. strong algorithms for image process area unit designed in step with the subsequent

three steps: first steps include, structure in pictures must be outlined as a applied mathematics model. Second, AN economical

improvement procedure to seek out sensible structures must be determined. we tend to advocate random improvement ways like

simulated tempering or settled variants of it that maximize the entropy whereas maintaining the approximation accuracy of the structure

live. Alternative improvement algorithms like interior purpose ways or continuation ways area unit equally appropriate. Third, a

validation procedure must take a look at the noise sensitivity of the discovered image structures. This three step strategy is incontestable

within the context of image analysis supported color and texture options. There has been a long-lived misunderstanding within the

literature of AI and uncertainty modeling, concerning the role of many-valued logics (and fuzzy logic). The revenant question is that of

the mathematical and pragmatic significance of a integrative calculus and also the validity of the excluded middle law. This confusion

even arises the first developments of even logic, even after several warnings of some philosophers of chance. This results in some level of

misunderstanding. It suggests that the basis of the controversies lies within the unfortunate confusion between degrees of belief and

what logicians decision "degrees of truth". The latter area unit typically integrative, whereas the previous can not be thus. It’s recalled

that any belief illustration wherever compositionality is taken as a right is absolute to at the worst collapse to a mathematician truth

assignment and at the best result in a poorly communicatory tool. we have a tendency to show the non-compositional belief illustration

embedded within the normal symbolic logic. It seems to be associate degree all-or-nothing version of chance theory.Computed

topography (CT) and resonance (MR) imaging area unit the foremost wide used photography techniques in designation, clinical studies

and treatment designing. This review provides details of automatic segmentation ways, specifically mentioned within the context of CT

and MR pictures. The motive is to debate the issues encountered in segmentation of CT and MR pictures, and also the relative deserves

and limitations of ways presently on the market for segmentation of medical pictures.

Keywords: Image processing and enhancement; Segmentation; Artificial intelligence techniques, computed tomography, magnetic

resonance imaging, medical images artifacts, segmentation

1. Introduction

Image Segmentation is that the strategy of partitioning a

digital image into multiple segments (sets of pixels, jointly

named as super pixels). The essential motive of

segmentation is to change and/or modification the

subsequent illustration of an image into one issue that is

simpler to analysis. Image segmentation is usually

accustomed notice objects and limits (lines, curves, etc.) in

photos. Plenty of precisely, image segmentation is that the

strategy of assignment a label to every component in a very

image such pixels with constant label share certain

characteristics.

2. Role of Segmentation

The main ways involve the segmentation of tomography

pictures and to get rid of the over segmentation drawback

with the assistance of varied cluster techniques .These

comprise of FCM cluster technique beside adaptive mean

shift. The process of image segmentation consists of

remodeling a picture into completely different phases (a

cartoon version), whereas keeping track of necessary

properties of every section. This could be used for analysis

of the image and/or for any process of the image, as every of

the various phases of the image is treated otherwise once a

segmentation method. One massive application space is

object detection. There exist many completely different

ways for image segmentation.

3. Problem with MRI

There are three problems that can be associated with MRI

images. The First problem is the partial volume effect in the

images. Due to presence of partial volume the segmentation

of the image may degrade. The second problem is the over

segmentation. The third problem is the Noise due to sensors

and related electronic system.

4. Methodology Used

The proposed algorithm is the modification of mean-shift

algorithm, termed as the adaptive mean shift algorithm. The

algorithm is divided into following layers whose functions

are different and their functionality is shown below along

with the respective figure. The basic steps are as follow:

Paper ID: SUB154820 2515





Figure 1: Proposed Algorithm

Step 1 Load Image: An image is a visual representation of

something. In this section it will load the brain image

followed by further functioning for the achievement of

segmentation.

Figure 2: Processing of an image

Step 2 Preprocessing: The terribly opening of study of brain

image segmentation is preprocessing. this pre-processing

includes two steps which are as follows:

•A: Homomorphic Low Pass Filtering– It provides

correction in intensity in-homogeneities and it performs

wavelet on image i.e. on original image by elimination

totally different variety of band say „HH‟ or „ll‟ or „HL‟ by

rotten the image up to bound level.

Figure 3: Preprocessing (Filtering)

•B: Image Stretching – It provides intensity values social

control across input channels supported dark and bright

points. It limits to a particular level of stretch image and

alter image across intensity value of the image.

Figure 4: Preprocessing (stretched image)

Step 3 Feature Vector: Feature vector area unit extracted per

input voxel. Intensity yet as spatial options i.e. voxel

coordinates is enlisted. Voxels coordinates for overall spatial

property of n is employed wherever n is that the variety of

input intensity channels. The posture of feature vectors is

input to the adaptive mean shift clustering stage. It handles

the stretch image across the rows as well as columns for

obtaining coordinate value. A pop up message will be

displayed as done as shown below:

Figure 5: Extraction of input voxels

Step 4 Adaptive Mean Shift:. Adaptive mean shift pixel

cluster implements the classic mean shift clustering

algorithm and it produces output pixels for clustered image

and output for averaged mean shift. The process starts by

clustering the input feature vectors, which represent the

multimodal MRI brain data using the FAMS implementation






of the AMS algorithm. The result obtained is further

classified into number of classes for further segmentation.

Figure 6: Output of Mean Shift

Step 5 Classes: we will calculate the length of cluster and is

divided by number of cluster and total number of cluster is

obtained.

Figure 7: Clustered image

Step 6 Fuzzy Rules: This section includes calculating the

distance between one cluster to another and further

normalizing the values to be obtained in fuzzy inference

system and rules are further created by making different

membership function and different values as per the need.

Different membership function includes say Triangular MF,

Trapezoidal MF, Gaussian MF etc.

Step7 Pruning: The number of mode is large as compared to

the initial data so mode pruning is therefore required. In

purning mechanism, we will calculate fuzzy inference

calculations and those obtained value of pixels which is less

than 0.5 is merged for pruning to avoid under segmentation.

Figure 8: Pruning on image

Step7 Fuzzy C Means (FCM): The cluster is performed on

numerous segmented elements and take a look at to beat the

over segmentation. Fuzzy c-means (FCM) could be a

knowledge cluster technique whereby every datum belongs

to a cluster to some extent that\'s mere by a membership

grade. It provides a technique that shows the way to cluster

knowledge points that populate some three-dimensional area

into a particular range of various clusters. Fuzzy Logic tool

case command perform fcm starts with associate initial

guess for the cluster centers, that area unit meant to mark the

mean location of every cluster. The initial guess for these

cluster centers is presumably incorrect. in addition, fcm

assigns each datum a membership grade for every cluster.

By iteratively change the cluster centers and also the

membership grades for every datum, fcm iteratively moves

the cluster centers to the proper location inside a knowledge

set. This iteration relies on minimizing associate objective

perform that represents the gap from any given datum to a

cluster center weighted by that knowledge point's

membership grade. The command performs fcm outputs a

listing of cluster centers and several other membership

grades for every datum. you\'ll use the knowledge came back

by fcm to assist you build a fuzzy logical thinking system by

making membership functions to represent the fuzzy

qualities of every cluster.

5. Result

The following method shows the comparison of k-mean and

fuzzy c-mean method and determines the numerical values

of different matter say gray matter, white matter, and

cerebro spinal fluid. It shows how under segmentation is

avoided using fcm and by comparing it will be more useful.

Figure 9: comparison of k-means and fcm






Figure 10: Comparison of Pruning values

The fig9 shows comparison of k means and fcm and shows

the calculated value through pruning .It shows value that is

obtained through mahalanobis distance i.e.(3215) and

according to fcm obtained value is i.e. (350) .hence

segmentation is carried out in three forms as shown below:

Figure 11: shows obtained result for selected MRI

6. Conclusion and Future Work

In this present thesis Neural Network approach using fuzzy

c-means clustering has been applied for segmentation of

MRI images. The work has been carried out in following

steps which is discussed above. The present work has been

done for improving segmented part in MRI image detection.

Further the present work can be extended to recognize

various clustering and the algorithm can be used for the

color image segmentation. The present work has been

carried to measure the segmentation through soft clustering

methods.

References

[1] H. Greenspan, A. Ruf, J. Goldberger Constrained

Gaussian mixture model framework for automatic

segmentation of MR brain images IEEE Trans Med

Imag, 25 (9) (2006), pp. 1233–1245.

[2] A.W.C. Liew, H. YanAn adaptive spatial fuzzy

clustering algorithm for 3D MR image segmentation

IEEE Trans Med Image, 22 (9) (2003), pp. 1063–1074.

[3] Cohen L., Cohen I., Finite-element methods for active

contour models and balloons for 2-d and 3-d images.

Pattern Anal Mach Intell IEEE Trans, 1993;

15(11):1131–1147.

[4] Abrantes AJ, Marques JS. A class of constrained

clustering algorithms for object boundary extraction,

IEEE Transactions on Image Processing, 1996; 5(11):

1507 – 1521.

[5] Vovk U, Pernus F, Likar B. A review of methods for

correction of intensity in homogeneity in MRI. IEEE

Trans Med Image, 2007; 26:405–21.

[6] Li CM, Liu JD, Fox M. 2005 “Segmentation of edge

preserving gradient vector flow: an approach toward

automatically initializing and splitting of snakes,” 2005

IEEE computer society conference on Computer vision

and pattern recognition. vol 1. p. 162–167.

[7] Xue JH, Pizurica A, Philips W, et al. An integrated

method of adaptive enhancement for unsupervised

segmentation of MRI bins images. Pattern Recogn Lett

2003; 24:2549–60.

Author Profile

Sana Tak received the B.E. From CSVTU Bhilai

(C.G.), India in Computer Science & Engineering in

the year 2013. She is currently pursuing M.Tech.

Degree in Computer Science Engineering with

specialization in Computer. Technology from CSVTU Bhilai

(C.G.), India. Her research areas include Software Engineering,

Neural Network, Pattern Recognition, Image Processing etc.

Toran Verma is Assistant Professor of Computer

Science & Engineering at Rungta College of

Engineering & Technology, Bhilai (C.G.), and India

as well as M.Tech. Coordinator (CSE) and has given

more than 10 years in RCET Bhilai, Completed

M.Tech. From CSVTU, Bhilai in 2011, Gold medalist as well as

University topper. More than 16 papers published in reputed

journals. Areas of Interest are Image Processing, Fuzzy logic and

Neural Network.


ameliorating brain image segmentation using fuzzy ... · ameliorating brain image segmentation...

Documents